Diversifying Neural Dialogue Generation via Negative Distillation

TL;DR

This paper introduces negative distillation, a novel training method for neural dialogue models that effectively reduces generic responses by using a negative teacher model and multi-level negative knowledge, improving response diversity.

Contribution

The paper proposes negative distillation, a new negative training paradigm utilizing a negative teacher model and multi-level negative knowledge to enhance dialogue response diversity.

Findings

Outperforms previous negative training methods significantly

Effectively reduces generic responses in dialogue generation

Improves diversity of generated responses

Abstract

Generative dialogue models suffer badly from the generic response problem, limiting their applications to a few toy scenarios. Recently, an interesting approach, namely negative training, has been proposed to alleviate this problem by reminding the model not to generate high-frequency responses during training. However, its performance is hindered by two issues, ignoring low-frequency but generic responses and bringing low-frequency but meaningless responses. In this paper, we propose a novel negative training paradigm, called negative distillation, to keep the model away from the undesirable generic responses while avoiding the above problems. First, we introduce a negative teacher model that can produce query-wise generic responses, and then the student model is required to maximize the distance with multi-level negative knowledge. Empirical results show that our method outperforms previous negative training methods significantly.